Back to Portfolio

About Me

Software Engineer, AI Developer, and Cybersecurity Undergrad based in Delhi, India.

Overview

I am a Computer Science undergraduate at the Faculty of Technology, University of Delhi (Class of 2028), specializing in Blockchain, Cybersecurity, and Cryptography.

My work centers on building dependable, production-ready software. I focus on developing full-stack web applications, autonomous AI workflows, and practical security systems. I prefer writing clean, maintainable code with clear deterministic logic over adding unnecessary architectural complexity.

AI & Software Engineering

In my current role as a Software Engineer Intern at Nowlez (AI Munshi), I work across AI systems and backend infrastructure:

  • Document Drafting Engines: Designed isolated, sandbox-based execution environments that allow AI models (Gemini) to safely generate, edit, and convert complex legal documents into Word and PDF formats.
  • Asynchronous Agent Architectures: Built background research subagents with isolated worker slots and webhook-based notification loops, keeping the user interface completely unblocked during long-running tasks.
  • Event-Driven Infrastructure: Deployed self-hosted LangGraph Agent Servers connected via webhooks to handle real-time chat turns, multi-message states, and human-in-the-loop confirmations.
  • Unified Authentication: Migrated fragmented authentication systems into a centralized Supabase Auth and PostgreSQL architecture with enforced dual email and phone verification.

Key Projects

AKTU DAA RAG System

A hybrid Retrieval-Augmented Generation (RAG) platform built for algorithm analysis coursework. It incorporates a dual-engine query classifier, dynamic regex-guided chunking, and Reciprocal Rank Fusion (fusing dense vector search and sparse BM25) to accurately answer student queries without hallucinations. Included 9 deterministic Python solvers to ensure mathematical precision on complex dynamic programming questions.

Project Eligify

An automated academic eligibility and marksheet verification platform. Built a multi-stage fallback parsing pipeline using PyPDF2, Tesseract OCR, and OpenCV image preprocessing to reliably extract scores and percentages from non-standard CBSE, state-board, and university marksheets.

DeepFake Audio Detection

An event-driven pipeline that screens forwarded voice notes for synthetic speech using a 3-signal ensemble (AASIST graph-attention network, wav2vec2 embeddings, and audio metadata heuristics), running inference efficiently on CPU. Secured 3rd place among 50+ teams in the IEEE AI War competition.

Systems & Security Research

During a summer research internship at SoCSE, RV University, I researched lightweight cryptography for Internet of Things (IoT) networks:

  • Designed SecureRPL, a lightweight security extension for the standard IETF RPL routing protocol (RFC 6550), integrating ChaCha20-Poly1305 authenticated encryption, replay protection, and trust-based routing into an embedded-friendly footprint.
  • Simulated multi-node mesh topologies in the NS-3 network simulator, benchmarking mitigations against sinkhole, replay, and node-capture attacks.

Community & Leadership

I led the organizing team for CityJS New Delhi at the Faculty of Technology, coordinating with 20+ speakers, sponsors, and partners while managing a 15-person volunteer team to deliver a 200-attendee tech conference.

Engineering Philosophy

I value simplicity, deterministic correctness, and security by design. When building software, I focus on understanding the underlying constraints before choosing the tools—whether that means replacing probabilistic LLM outputs with precise algorithms or building resilient backends with solid fallback strategies.